Hi,
On Sat, Jan 12, 2019 at 07:24:59PM +0100, Sebastien Marie wrote:
> The lxml test calls in loop, and in several threads, XML parsing on
> errornous documents and check the error. The test dies because free(3)
> detects memory corruption and call abort(3).
>
> Without the option --without-threads on libxml, the lxml's test passes.
>
> I checked on amd64 that no symbols were added/removed on libxml2.so, and
> WANTLIB is still the same too (pthread was already here). The testsuite
> of libxml still pass and the one from py-lxml passes (whereas it wasn't
> the case before).
>
> I would be interested to know any downside of enabling thread safety on
> libxml2.
I'd say, just let's try it. The --without-threads had been added a long
time before we got rthreads, and it may be that nobody since then tried
to remove it:
----------------------------
revision 1.67
date: 2003/08/26 16:33:11; author: marcm; state: Exp; lines: +3 -2;
branches: 1.67.2;
Update again to 2.5.10 and this time *disable threading*. Reports are
that this fixes a whole range of breakage with other ports that were
confused by the threading option.
ok naddy@
=============================================================================
Ciao,
Kili
